DATING SCAM DETECTORFor years, the single most reliable check in this entire guide library has been simple: ask for a live, unscripted video call. AI voice cloning and deepfake video technology are starting to complicate that advice, and it's worth understanding exactly how, rather than either panicking or dismissing it.
What's already documented — and it's not romance scams (yet)
The clearest, most financially damaging AI-cloning cases documented so far are "distress" scams impersonating a family member in an emergency, not romantic partners. Security researchers have found that as little as three seconds of audio — pulled from a public social media video or even a voicemail greeting — is enough to clone a convincing version of someone's voice. In one widely reported 2024 case, a fraudulent transfer of $25.6 million was carried out using this kind of impersonation, and industry tracking suggests victims lost more than $5 million to these "distress" scams in 2025 alone, with roughly one in three people who engage with an AI-cloned voice call ending up losing money, averaging over $18,000 per loss. The mechanism is worth understanding because it's directly transferable to romance scams: some of these scams work specifically because when a suspicious recipient calls back through the same (compromised) account or number, the scammer answers using the cloned voice, defeating the most obvious verification step.
Why this matters for online dating specifically
Security researchers and fraud-prevention companies have been explicit that this same technology is expected to migrate into romance scams as it becomes cheaper and more accessible — a scammer who has scraped a stolen photo set can, in principle, pair it with a cloned or AI-generated voice to get past a phone call, and increasingly, real-time deepfake video filters are being demonstrated that can alter a face during an actual video call rather than requiring pre-recorded footage. As of now, documented cases specifically inside romance scams remain less common than family-impersonation distress scams, but the underlying tools are the same, and researchers broadly expect this gap to close.
How to adapt the video-call test
A video call is still meaningfully harder to fake convincingly in real time than a voice call or a set of photos, but it's worth making it more specific rather than treating any video as sufficient proof. Ask for small, unscripted, real-time actions during the call — turn your head to the side, hold up a specific number of fingers, or repeat a random word you just picked. Real-time deepfake filters currently struggle most with fast, unexpected movement and profile angles, which is exactly why asking for them is more useful than simply watching someone talk head-on.
Other practical adjustments
Be more cautious than before about a phone call alone being sufficient verification, particularly if it comes through a messaging app rather than a number you already have confirmed for that person. If something about a call feels slightly "off" — a strange audio quality, a face that doesn't quite track natural movement, delayed or generic responses to specific personal questions — treat that instinct as worth acting on rather than dismissing, even if you can't articulate exactly what's wrong. And apply the same standing rule that protects against every version of this fraud: any financial request, regardless of how it's verified, gets discussed with someone outside the relationship before money moves.
How regulators are responding
Financial and consumer-protection regulators have started treating AI-enabled impersonation as its own tracked category rather than folding it into generic fraud. The FTC has issued public consumer warnings specifically about AI voice-cloning scams, and several U.S. states have passed or proposed disclosure requirements for AI-generated audio and video used in fraudulent contexts. Enforcement is still catching up to the technology — the practical protection right now is behavioral (the real-time verification steps above), not regulatory.
Why this is likely to get more common, not less
The tools behind voice cloning and deepfake video have gone from requiring specialized technical skill to being available as consumer-facing apps and services in just a few years, with cost dropping accordingly. Security researchers broadly expect the same trajectory that played out with photo-based catfishing — cheap, accessible tools spreading a tactic from a small number of sophisticated operations to widespread, commodity use — to repeat with voice and video.
This doesn't replace the fundamentals
AI tools change the mechanics of impersonation, but they don't change the underlying pattern this entire site is built to catch — urgency, a financial ask, and reluctance to be verified in ways that are hard to fake. A cloned voice still needs a stolen photo set, a script, and a plausible reason for money to change hands. Running the actual conversation through our free checker still catches the parts of the scam that AI hasn't changed at all.
